Basler ToF camera ROS driver

This is a ROS driver for the Basler ToF ES (Engineering Sample) 3D camera:

Installation

• Get the file basler-tof-driver-1.4.1.1450-x86_64.tar.gz (available from baslerweb.com) and unpack it. (This driver has been tested with versions 1.0.6, 1.1.0, 1.2.1, 1.3.1 and 1.3.2 of basler-tof-driver, but should work with any version >= 1.0.6 . Version 1.2.1 or later is recommended due to the improvements in the postprocessing filters.
• When you upgrade from one BaslerToF library version to another, make sure to remove your build directory and rebuild. For some reason, cmake does not properly detect that the node has to be rebuilt.
• Move the directory BaslerToF from that archive to /opt/BaslerToF.
• Do not put this directory on your LD_LIBRARY_PATH; it contains custom versions of Qt and other libraries that interfere with the system libraries.
• Clone this repo into your Catkin workspace and install as usual. If unsure, refer to the commands in the .travis.yml file.

Running

roslaunch basler_tof basler_tof.launch

ROS API (basler_tof_node)

Published Topics

confidence/image_raw (sensor_msgs/Image)

• The confidence image (each pixel represents the confidence in the depth value). This is the same as the intensity image, but with too bright / dark pixels beyond a certain threshold replaced by NaN values.

confidence/camera_info (sensor_msgs/CameraInfo)

• The corresponding camera info (calibration values).

---

depth/image_raw (sensor_msgs/Image)

• The depth image.

depth/camera_info (sensor_msgs/CameraInfo)

• The corresponding camera info (calibration values).

---

intensity/image_raw (sensor_msgs/Image)

• The intensity image.

intensity/camera_info (sensor_msgs/CameraInfo)

• The corresponding camera info (calibration values).

---

points (sensor_msgs/PointCloud2)

• Point cloud, directly from the device driver. Has intensities, but cannot be user-calibrated.

---

depth/points (sensor_msgs/PointCloud2)

• Point cloud computed from the depth/image_raw and depth/camera_info topic. No intensities, but takes into account user calibration. This topic is not directly published by the node, but by the depth_image_proc/point_cloud_xyz nodelet in the launch file.

There are also several rectified topics published via image_proc in the launch file, for example intensity/image_rect:

Parameters

Not Dynamically Reconfigurable Parameters

~frame_id (string, default: "camera_optical_frame")

• The frame id used in the sensor messages.

~camera_info_url (string, default: None)

• An URL pointing to a camera info yaml in camera_info_manager format.

Dynamically Reconfigurable Parameters

See the dynamic_reconfigure package for details on dynamically reconfigurable parameters.

~frame_rate (int, default: 15)

• Acquisition rate at which the frames are captured. [Hz] Range: 1 to 15

~exposure_auto (bool, default: True)

• Enable automatic exposure mode.

~exposure_time (int, default: 15000)

• Exposure time when exposure_auto = false. Range: 300 to 30000

~exposure_agility (double, default: 0.4)

• Reaction speed of the continuous exposure mechanism when exposure_auto = true. Range: 0.1 to 1.0

~exposure_delay (int, default: 2)

• The number of frames it takes until the continuous exposure mechanism reacts to a change in the exposure time when exposure_auto = true. Range: 1 to 10

~confidence_threshold (int, default: 2048)

• Pixels have to exceed the confidence value specified here in order to be taken into account for the distance measurement Range: 0 to 65536

~spatial_filter (bool, default: True)

• Enable the spatial noise filter (uses the values of neighboring pixels to filter out noise in an image).

~temporal_filter (bool, default: True)

• Enable the temporal noise filter (uses the values of the same pixel at different points in time to filter out noise in an image).

~temporal_strength (int, default: 190)

• Strength of the temporal filter. The higher the value, the further back the memory of the filter reaches. High values can cause motion artifacts, while low values reduce the efficacy of the filter. Range: 50 to 240

~outlier_tolerance (int, default: 2000)

• Maximum distance that a pixel's depth is allowed to differ from that of its neighboring pixels [mm]. Range: 0 to 65535

Calibration

Due to a bug in camera_calibration (ros-perception/image_pipeline#246), cameracalibrator.py does not work well with mono16 image encodings. To work around this, uncomment the following two lines in src/basler_tof_node.cpp and re-run catkin_make:

  // intensity_cvimg.image.convertTo(intensity_cvimg.image, CV_8U, 1.0 / 256.0);
  // intensity_cvimg.encoding = sensor_msgs::image_encodings::MONO8;

After this, run calibration as usual:

rosrun camera_calibration cameracalibrator.py image:=/camera/intensity/image_raw camera:=/camera/intensity --size 6x7 --square 0.05

Troubleshooting

Q: I get the following error:

Failed to load the producer file ProducerTOF.cti: ProducerTOF.cti: cannot open shared object file: No such file or directory

A: This means that the file ProducerTOF.cti (from BaslerToF) is not on the LD_LIBRARY_PATH. Make sure that you have installed the basler-tof-driver to /opt/BaslerToF or adjust the basler_tof_path argument of basler_tof.launch accordingly.

Open Issues

According to the Basler documentation, the point cloud's origin is at the front of the camera housing:

The reference location used is Anchor. This is the default value. The reference point is fixed to the origin of the camera's coordinate system, i.e. x and y are 0 at the central pixel of the sensor and z is 0 at the front of the camera housing.

However, the camera_optical_frame is defined by the camera's lens, which could result in an offset in the z direction of several millimeters.

Sensor Sample

A sample output of the sensor can be found in doc/sensor_samples/basler_tof.bag.

Build Status

Build Job Status (Travis)

	Indigo	Jade	Kinetic
basler_tof			N/A

Note: There is a Kinetic branch, but Travis doesn't provide Xenial hosts that are required for building on Kinetic.

Doc Job Status (ROS build farm)

	Indigo	Jade	Kinetic
doc